Veo video watermark detector

Veo watermark

What the Veo video watermark detector checks

Google's Veo can attach two separate signals to its generated video: a visible mark that shifts position over the course of a clip, and SynthID, an invisible signal embedded in the video's frames. Checking for the visible mark is something our media workflow can do directly, scanning across the clip rather than trusting a single glance. Confirming SynthID with certainty is a different matter, since only Google's own detector can rule on that signal authoritatively.

Drop the clip into the box on this page and our media workflow scans multiple points across the video for the visible mark, reporting what it finds honestly rather than treating a single glance as the whole picture.

The problem, in plain terms

Why detecting Veo's visible mark isn't a one-glance check

A fixed corner logo is easy to check: look at that one corner and you have your answer for the whole clip. Veo's visible mark does not sit still, so a check based on a single frame or a single corner can miss it entirely if the mark happens to be positioned elsewhere at that exact moment.

SynthID is a separate matter entirely. It's not something you can look for the way you look for the visible mark, since it's a sub-perceptual signal that requires Google's own detector to confirm. Conflating the two checks is the most common mistake here.

How it works

How to detect Veo's visible video watermark in three steps

The scan runs through the upload control on this page.

Step 1

Upload the original clip rather than a re-share, since re-encoding can shift the mark's visibility.

Step 2

Our media workflow scans multiple points across the clip, since the mark's position moves and a single frame can miss it.

Step 3

Read the result as a statement about the visible mark specifically. It says nothing about SynthID either way.

Why it matters

What getting this wrong actually costs

Assuming a clip is unmarked because one frame looked clean is a mistake with real consequences: republishing generated content as though it were camera-original footage. Assuming that a clean-looking visible check also rules out SynthID is a second, separate mistake, since the two signals have nothing to do with each other. Both matter for anyone trying to evaluate a Veo clip honestly.

The mechanism

Why Veo has one signal you can check and one you can't

The visible mark exists specifically to be seen, which is why scanning through a clip can actually confirm or rule it out. SynthID exists specifically to not be seen, embedded sub-perceptually in the frame data, which is why no amount of looking at a clip will confirm or rule it out on its own. That asymmetry is a real feature of how the two systems are built, not an inconsistency in how this page explains them.

Straight answer

Can checking for the visible mark tell me anything about SynthID?

No. They're independent signals. A clip can show the visible mark clearly and still carry SynthID, or show no visible mark at all and still carry SynthID, since the two aren't linked. Only Google's own detector can confirm SynthID with certainty; this page checks the visible mark and states that limitation directly rather than glossing over it.

Treat a visible-mark check as answering exactly one question, not both.

Who uses this

Who checks for a Veo watermark, and when

A few recurring, legitimate reasons to check.

Editors and publishers

Confirming whether a clip they're about to use carries the visible mark before it goes into a public deliverable.

Moderators and fact-checkers

Assessing whether a circulating clip shows signs of Veo generation before drawing a conclusion.

Researchers

Understanding the difference between a signal they can check themselves and one that requires Google's own tool.